Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
System of polynomial equations wikipedia , lookup
Polynomial greatest common divisor wikipedia , lookup
Eisenstein's criterion wikipedia , lookup
Polynomial ring wikipedia , lookup
Factorization of polynomials over finite fields wikipedia , lookup
Factorization wikipedia , lookup
UNIVERSITY OF BUCHAREST FACULTY OF MATHEMATICS AND COMPUTER SCIENCE BACHELOR THESIS ALEXANDER POLYNOMIALS OF THREE MANIFOLDS CRISTINA ANA-MARIA ANGHEL DIRECTED BY: PROF. DANIEL MATEI PROF. VICTOR VULETESCU 20th JUNE 2013 Contents Acknowledgements iii iv Rezumat v Introduction 1 1 Alexander polynomials of groups 1.1 Alexander modules . . . . . . . 1.2 Alexander varieties . . . . . . . 1.3 Alexander polynomials . . . . . and . . . . . . . . . spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Knots and links in 3-manifolds 2.1 Knots and links complements: (co) homological properties . . 2.2 Knots and links complements: homotopical properties . . . . . 2.3 Links, planar diagrams and Reidemeister moves . . . . . . . . 2.4 Links, braids and Markov moves . . . . . . . . . . . . . . . . . 2.5 Alexander trick and Vogel algorithm . . . . . . . . . . . . . . 2.6 The fundamental group of the complement: braid group picture 2.7 The fundamental group of the complement: Wirtinger picture 3 Alexander polynomial of knots and links 3.1 Alexander polynomial via free Fox calculus . . . . 3.2 Alexander polynomial via braid groups . . . . . . 3.3 Alexander-Conway polynomial via skein relations 3.4 General properties of the Alexander polynomial . . . . . . . . . . . . . . . . . . . . . . i 4 4 6 6 8 8 9 10 10 14 18 19 23 . . . . . . . 23 . . . . . . . 25 . . . . . . . 28 . . . . . . . 31 4 Applications 4.1 Fox coloring . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 2-Bridge knots and links . . . . . . . . . . . . . . . . . . . . . 4.3 Metabelian quotients . . . . . . . . . . . . . . . . . . . . . . . 33 33 34 37 Appendix 42 A1. Manifolds and duality . . . . . . . . . . . . . . . . . . . . . . . 42 A2. Immersions, embeddings, isotopy . . . . . . . . . . . . . . . . . 44 A3. Rings and orders . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Directions for further study 48 Bibliography 49 ii Acknowledgements I would like to thank first of all to my advisors Prof. Daniel Matei and Prof. Victor Vuletescu for the nice subject, for patiently encouraged me along the way and for the support during the preparation of this work. I would like to express my gratitude to Prof. Matei for all help: many discutions, books, papers and hints he gave to me during these years. I acknowledge many things I learned from Prof. Vuletescu during these all 3 years at his classes at FMI and SNSB. I thanks a lot to Prof. Dorin Cheptea for his guidance, answers to many questions, suggestions and for reading a first version of this thesis. Also, I would like to thank to Prof. Dorin Cheptea and Prof. Sergiu Moroianu for so many things I learned from their course on braids at SNSB. I have reserved the last but not least word from these acknowledgements, for the person who has determined my math trajectory, how it looks like at this moment: Prof. Liviu Ornea. Along these years, his lectures were for me, a great scientific experience. His Algebraic Topology course, I think, was a key moment for the decision to follow a topological direction for my studies. iii ...in simple words... iv Rezumat Subiectul acestei lucrări ı̂l constituie teoria nodurilor. Acest domeniu a apărut odata cu lucrările lui Gauss legate de numere de ı̂nlănţuire. Ca ramură distinctă a topologiei s-a constituit la ı̂nceputul secolului XX prin lucrările lui Poincare, Alexander si Dehn. În anul 1928 este introdus pentru prima data aşa numitul polinom Alexander. Acest invariant, este suficient de puternic pentru a detecta diferenţe inaccesibile fără el, dar totodată relativ limitat: de exemplu nu poate detecta diferenţa dintre două noduri care sunt unul imaginea ı̂n oglinda a celuilalt. Acest neajuns este rezolvat parţial ı̂n anii ’90 odată cu apariţia polinomului Jones şi a ı̂ntregii pleiade de invarianţi care au apărut ulterior (invarianţi cuantici, invarianţi polinomiali in 2 variabile, etc). Scopul acestei lucrări este de a prezenta pe de o parte polinomul Alexander sub multiplele sale faţete iar pe de alta de a studia relaţia dintre varietaţile caracteristice şi reprezentările metabeliene ale grupului unui nod. Lucrarea este constituită din 4 capitole, o introducere şi un apendix destinat descrierii unor noţiuni şi teoreme necesare pe parcursul tezei. In cadrul introducerii este prezentat subiectul general al teoriei nodurilor şi metodele de abordare ale problemelor prin prisma teoriei Alexander. Primul capitol este destinat introducerii principalelor ingrediente utilizate de-a lungul tezei: modulul si polinomul Alexander. Totodată sunt introduse şi varietăţile caracteristice care vor juca un rol esenţial in cadrul ultimei părţi a lucrării. În cadrul capitolului 2 sunt prezentate o serie de rezultate cu privire la topologia complementelor de noduri. Astfel, primele două paragrafe se referă la proprietăţile omologice şi omotopice ale complementelor. În continuare sunt prezentate diagramele planare, mişcările Reidemeister, relaţia dintre linkuri şi braiduri, precum şi mişcările Markov. O atenţie deosebită este acordată ı̂n paragraful 5, metodelor de construcţie a braidurilor asociate v linkurilor: se demonstrează teorema Alexander şi se descrie algoritmul Vogel. Ultimele paragrafe ale acestui capitol se referă la grupul fundamental al complementului ı̂n cele două abordări uzuale: Wirtinger şi cea care utilizează grupul braid, mai precis scufundarea acestuia ı̂n grupul de automorfisme ale grupului liber. Următorul capitol se referă ı̂n exclusivitate la trei dintre modalităţile de construcţie ale polinomului Alexander: prin calcul Fox, via grupuri braid şi cu relaţii skein. Este descrisă reprezentarea Burau şi demonstraţia existenţei polinomului Alexander-Conway. Din păcate am omis descrierea contrucţiei care foloseşte suprafaţa Seifert. În ultimul paragraf sunt prezentate proprietăţile generale ale polinomului Alexander, inclusiv relaţiile Torres. Capitolul 4 este destinat aplicaţiilor. În prima parte (paragrafele 1 si 2), sunt prezentate colorările Fox, relaţia dintre colorări si proprietăţi aritmetice ale polinomului Alexander, precum şi o clasă specială de linkuri: cele cu 2 poduri (2-bridge links). Cu privire la acestea din urmă, este prezentat grupul fundamental şi proprietăţile speciale ale polinomului Alexander. În paragraful 3 se prezintă relaţia dintre varietăţile caracteristice şi anumite clase de reprezentări metabeliene. Rezultatele principale ale acestui paragraf sunt teoremele 4.3.8 şi 4.3.9 din [19] ı̂n care se calculează cardinalul morfismelor, epimorfismelor precum şi invarianţii Hall ı̂n cazul unor extinderi metaciclice. Finalul acestui paragraf este destinat unui rezultat similar, pentru cazul morfismelor de la grupul unui nod cu 2 poduri ı̂ntr-un grup diedral. Lucrarea se ı̂ncheie cu o scurtă descriere a trei potenţiale direcţii de continuare a studiului ı̂nceput ı̂n această teză. vi Introduction The main objects we are talking about in this thesis are (oriented) knots K and links L in R3 or S3 . Below are three examples: the figure eight knot, the Whitehead link and the Borromean link (all pictures use the [28] on-line converter). We will deal here only with polygonal or smooth links and it is known the following: Theorem Every smooth link is isotopic with a polygonal one. Two links are considered the same (equivalent), denoted by L ' L0 if they are ambient isotopic. An equivalent assumption is that there exists a homeomorphism of S3 , which preserves the orientation and takes L to L0 cf. [16] . For smooth links, ambient isotopy is the same as isotopy cf. A2. A coarser equivalence relation is: Definition Two links are weakly-equivalent, denoted by L ∼ L0 if there is a homeomorphism of S3 taking one in the other. The difference is fundamental. For example the two trefoils 1 are weakly equivalent but not equivalent. The aim of knots/links theory is to classify them up to (weak)-equivalence. A surprisingly quite recent result is the following theorem due to Gordon and Luecke. It appeared first in [8] and with full details in [9]: Theorem Two knots with homeomorphic complements are weakly equivalent. Remark First of all the two trefoils have homeo complements but they are not equivalent. For example the Jones polynomial distinguishes them. Secondly, the above theorem is not true for links ! However, even if the weak-equivalence for knots is considered, the homeo type of the complement is a hardly tractable invariant. It is very desirable to have weaker invariants which are easy to compute, possibly by combinatorial techniques. One such invariant is the fundamental group of the complement: π = π1 (S3 \K). Although π distinguishes the unknot, it is not a complete invariant for weak-equivalence. For example, the square and granny knots have the same π but are not weak-equivalent. In fact the peripheral structure distinguishes them. 2 π is non-abelian except when the knot is trivial and usually given via a presentation by generators and relations. It is a hard problem to decide when two presentations give isomorphic groups. The next step toward a computable invariant is the so-called Alexander module M of the knot group. For a presentation of the group, there are several methods to produce presentations for this module. Then, using Fitting ideals, the (computable) Alexander polynomial invariant is obtained. 3 Chapter 1 Alexander polynomials of groups and spaces The aim of this chapter is to associate in topological and algebraic contexts tractable invariants: modules, varieties, polynomials. The main references are [18], [4], [21] and [25]. 1.1 Alexander modules Let X a connected CW-complex of finite type, with only one 0-cell x0 and π = π1 (X, x0 ). Let H = πab = H1 (X, Z). In some situations we can use complex coefficients instead of integers. From the standard correspondence between coverings of X and subgroups in π, associated to the surjection π → H there is a covering p : X̃ → X named the maximal abelian covering. In such a situation the homology of the covering became a module over the group of deck transformation, wich, in this setting, is nothing else than H. This H-module structure on H1 (X̃, Z) became a true module structure over ZH i. e. the group ring of H. We summarize with the following: Definition 1.1.1 The ZH- module Bπ := H1 (X̃, Z) is named the Alexander invariant of X. Definition 1.1.2 The ZH- module Aπ := H1 (X̃, p−1 (x0 )Z) is named the Alexander module of X. 4 The previous structure can be considered in a purely algebraic setting. Let G be a group, G0 its commutator and Gab its abelianisation. Let G00 the commutator of G0 . Using the exact sequences: 0 → G0 → G → Gab → 0 0→ G0 G00 → G G00 → Gab → 0 it can be proved the following: Lemma 1.1.3 The pseudo-action by conjugation of Gab on G0 induce a 0 well-defined action on GG00 . In the lemma above the term pseudo-action is used because the conjugation is not well-defined on G0 . It becomes well-defined only after factorization. As in the topological case, by passing to the group ring we obtain a ZGab 0 module structure on GG00 . In this algebraic setting it is called the Alexander invariant of G and is denoted by BG [4]. In the case where G = π1 (X) both constructions give the same answer, i.e. with the notations above and from the previous section, we have the following identification between the topological and algebraic Alexander invariant of X respective π [25]: Theorem 1.1.4 There is a natural isomorphism of ZGab - modules G0 G00 ' H1 (X̃, Z). Also, the homology sequence for the pair (X̃, p−1 (x0 )) gives 0 → H1 (X̃) → H1 (X̃, p−1 (x0 )) → H0 (p−1 (x0 )) → Z, the last map being the augmentation : H0 (p−1 (x0 )) → H0 (X̃) ' Z. In Alexander-type terms, we obtain the following exact sequence, 0 → BG → AG → I → 0 where I is the kernel of the evaluation map : ZGab → Z. If we identify ZGab with Λ = Z[t1 ±1 , ..., tq ±1 ], then I = (t1 − 1, ...tq − 1) and AG = ZGab ⊗ I. 5 1.2 Alexander varieties For X, G as above, we have the following: Definition 1.2.1 For k ≥ 1 the k th Alexander (characteristic) variety of X (or G) is the (k + 1)th support variety of the Alexander module: Vk (G) = Vk+1 (AG ⊗ C) ∪ {I}, where I is the identity (trivial representation) in the character torus T. From the definition above, it is clear that the characteristic varieties form a decreasing sequence (as the E 0 s form an ascending one cf. appendix A3) and that they depend only on GG00 . Also we have the following important remark which relates the support varieties of various levels of the Alexander module with to those of the Alexander invariant: Remark Vk+1 (AG ⊗C) = V (Ek (AG ⊗C)) = V (Ek−1 (BG ⊗C)) = Vk (BG ⊗C). 1.3 Alexander polynomials For X and G as above, with q = b1 (X) ≥ 1 we have: Definition 1.3.1 The Alexander polynomial of X (or G) is ∆X = ∆G = ∆1 (AG ). Remark It depends only on G (in fact only on multiplication by units in Λ. G ) G00 and is defined up to For a group G, the deficiency, def (G), is the minimum of the difference between the number of generators and relations, over all presentation. The following theorem from [5] expresses the first elementary ideal of the Alexander module, in terms of the Alexander polynomial when the deficiency is positive: Theorem (Eisenbud-Neumann) 1.3.2 If b1 (X) = 1 then E1 (∆G ) = (∆G ) (i.e. it is a principal ideal). If b1 (X) ≥ 2 and def (G) > 0 then E1 (AG ) = I · (∆G ). The next result shows the relation between the Alexander polynomial and the characteristic varieties: 6 Proposition 1.3.3 ∆G = 0 iff V1 (G) = TG . If ∆G 6= 0 then V (∆G ) if q > 1 W1 (G) = V (∆G ) ∪ {I} if q = 1 If q ≥ 2 then W1 (G) = ∅ iff ∆G is constant. Corollary 1.3.4 If def (G) > 0 then V1 (G) = V (∆G ) ∪ {I}. 7 Chapter 2 Knots and links in 3-manifolds In the following K and L denote a knot or a link in S3 . X is the complement. For L we shall denote by K1 , ..., Km the knot components of L. N is a tubular neighborhood of L; it is the union of m closed solid tori. E = S3 \Int(N ). Obviously E is a 3-manifold with boundary, X is an open 3-manifold, they have the same homotopy type and consequently the same π1 and the same homology. In fact E is a deformation retract of X. 2.1 Knots and links complements: (co) homological properties The next two theorems describe the (co)homological structure of the link complement. They show that the homological information is totally insensitive to knotting. Among the main references there are [22] and [6]. Theorem 2.1.1 H0 (X) = Z, H1 (X) = Zm , H2 (X) = Zm−1 and Hi (X) = 0 for i ≥ 3. On the cohomological side we have: Theorem 2.1.2 H 0 (X) = Z, H 1 (X) = Zm , H 2 (X) = Zm−1 and H i (X) = 0 for i ≥ 3. 8 2.2 Knots and links complements: homotopical properties Definition 2.2.1 For a group π and a positive integer n by K(π, n) we denote the unique homotopy type of spaces (called Eilenberg-Maclane) for which the only nonzero homotopy group is π in dimension n. From the homotopical point of view, the first main result is [2]: Theorem 2.2.2 For a knot, the complement is an Eilenberg-Maclane K(π, 1) space. For links, with the following: Definition 2.2.3 A link L is split if it can be decomposed as L1 ∪ L2 with the L0i s in the interior of disjoint 3-balls. we have: Theorem 2.2.4 For a link, the complement is an Eilenberg-Maclane K(π, 1) space iff L is not split. If L = L1 ∪...∪Lk with the L0i s nonsplit, then the complement X is homotopic equivalent with K(π, 1) ∨ S2 ∨ ... ∨ S2 , where the number of spheres is k − 1. We now turn to the peripheral system and semi-direct product structure of knot/link groups. First of all we want to mention the following theorems: Theorem 2.2.5 A knot is trivial iff π1 is infinite cyclic. Theorem 2.2.6 A knot is nontrivial iff π1 (∂V ) → π1 (X) is injective. From the previous theorem every nontrivial (different from Z) knot group has a specified subgroup isomorphic to Z2 . This subgroup is named the peripheral structure on π. It is defined only up to conjugation. A celebrated theorem of Waldhausen [2] is the following: Theorem 2.2.7 A knot is determined up to equivalence by the isomorphism class of its peripheral structure. For example [25], Fox used this theorem to prove that the square and granny knots, even having the same π, are distinct. Another important fact about knot groups is that the surjection π → πab has a section which sends the generator (i. e. the homology class of a meridian) into its homotopy class. It follows that the knot group is in fact a semidirect product π 0 o πab . 9 2.3 Links, planar diagrams and Reidemeister moves In this section we will explain the representation of links by planar diagrams. Choosing a generic plane in the 3-space and projecting the link on it we obtain a figure consisting of a number of crossings. We should always consider projections with only transverse crossings. Of course, the number of crossings or the succession of their type along a component of the link are by no means invariants of the link. One major problem concerning the projections is to decide when two of them represent the same link. The following type of moves called Reidemeister moves do not change the isotopy type of the link: The main point is that we have the following remarkable theorem [23]: Theorem 2.3.1 Two planar link diagrams represents equivalent links iff they can be transformed one in the other by Reidemeister moves. 2.4 Links, braids and Markov moves This section is devoted to braids and their relations with links cf. [1] [22]. 10 Definition 2.4.1 A n-braid consist of: 1. n points in R3 with the z-coordinate a and the x coordinate strictly increasing denoted by Pi 2. n points in R3 with the z-coordinate b and the x coordinate strictly increasing denoted by Qi 3. A permutation and for every i a path from Pi to Q(i) , such that on every path the z-coordinate is strictly decreasing. 3. a < b and the paths are disjoints. In the picture below, taken from [22] there are 2 examples of 3-braids: As for links, there is a notion of equivalent braids up to isotopy; the conditions 1 − 4 must be verified at each moment and the the vertices are fixed through the isotopy. Even if at first sight some conditions from above might be redundant, it is not the case. For example, the following two braids from [22] are not isotopic: 11 The operation of gluing two n-braids define a group structure on the classes of equivalent braids: the Artin braid group Bn . A celebrated theorem of Artin asserts that if we denote by σi the following braid (picture from [22]) then: Theorem 2.4.2 The σi ’s for i = 1...n − 1 are a generator system for Bn with the relations: σi σj = σj σi for | i − j |≥ 2 σi σi+1 σi = σi+1 σi σi+1 for 1 ≤ i ≤ n − 2. For example, in the group Bn the inverse of σi is simply Another important result due also to Artin is the representation of Bn in the automorphisms group of Fn , the free groups on n letters x1 , ..., xn . Let ξi the autmorphism of Fn defined as follows: ξi (xi ) = xi xi+1 xi −1 ξi (xi+1 ) = xi ξi (xj ) = xj for j 6= i, i + 1. 12 Artin representation theorem asserts that: Theorem 2.4.3 1. The map ϕ : Bn → Aut(Fn ) defined on generators by: ϕ(σi ) = ξi is a well defined injective morphism from Bn to Aut(Fn ). 2. An element ξ ∈ Aut(Fn ) is in the image of ϕ iff there are words Ai ∈ Fn and a permutation such that: ξ(xi ) = Ai xε(i) Ai −1 for 1 ≤ i ≤ n ξ(x1 ...xn ) = x1 ...xn . Proof: The main point is to define for each σ ∈ Bn an explicit automorphism σ̄ of Fn . We consider a point P in the z = a plane with the x-coordinate smaller than any x-coordinate of the Pi ’s and also that its projection Q on the z = b plane has the same property in relation with the Qi ’s. Consider the ambient space R3 with the strings of σ removed. The plane z = a became a plane with n-points removed, having hence the fundamental group Fn . We can think the generators xi ’s as loops around the Pi ’s based at P . We push down each xi to a loop in the z = b plane based at Q. Because the z = b plane with the Qi ’s removed has the same fundamental group Fn , this push down operation is in fact a map σ̄ : Fn → Fn . The theorem is proved along the followings small steps: 1. if l1 and l2 are P based loops in the pointed z = a plane, then σ̄l1 and σ̄l2 are homotopic loop in the pointed z = b plane; 2. if l1 and l2 are P based loops in the pointed z = a plane, then σ̄(l1 l2 ) = σ̄l1 σ̄l2 ; 3. σ̄ is bijective; in fact it has as inverse the push up operation; From the above steps σ̄ is an automorphism. 4. if τ is another braid isotopic with σ then σ̄ = τ̄ (in fact, according to the next step it is sufficient to prove this for a braid isotopic with the trivial one); 5. στ ¯ = σ̄τ̄ ; 6. σ̄(x1 ...xn ) = x1 ...xn ; 7. for σ = σi , σ̄i verify σ̄i (xi ) = xi xi+1 xi −1 , σ̄i (xi+1 ) = xi and σ̄i (xj ) = xj for j 6= i, i + 1; 8. as Bn =< σ1 , ..., σn > and σ̄ is an automorphism of Fn then σ̄xi has the form Ai xε(i) Ai −1 for Ai ’s words in Fn and ε a permutation in Sn (induction on the length of σ as word in the σi ’s and their inverses). It remains to show that a morphism of the above form is induced by a braid. 13 The proof is by induction on l = Σl(Ai ), the sum of the lengths of the Ai ’s in Bn . For l = 1 there is nothing to prove: ξ is the identity. Suppose the assertion true for l < m and let a ξ with l = m. By multiplication of all the relations ξ(xi ) = Ai xε(i) Ai −1 , the left hand side has length n and so some cancellations must occur on the right. A careful analysis of this process (cf. [22] pp. 90) together with the inductive hypothesis finishes the proof. QED In the end of this section we discuss the relation between links and braids. First of all there is a natural operation of ”closing” a braid, producing a link. A fundamental result due to Alexander asserts the converse: any link is the closure of a braid. We will talk about it in the next section. An important question is when two braids give the same isotopy class of links. As in the case of planar representation of links, there are two type of moves on braids which leave the associated link invariant: 1. the first is simply conjugation by another braid 2. for the second, if the initial braid b is in Bn , we shall ebbed b in Bn+1 by adding one string and the move is simply multiplication by σn±1 . The following theorem due to Markov, gives the complete answer to the above question. Theorem 2.4.4 Two braids give equivalent links iff they are related by a finite number of Markov moves. 2.5 Alexander trick and Vogel algorithm The main references for this section are Prasolov and Sossinsky [23], Kassel and Turaev [14] and Manturov [17]. It is devoted to Alexander theorem below; the first proof we present uses the so called Alexander trick. The second consists of the Vogel algorithm. Theorem 2.5.1 Any link is the closure of a braid. Proof: Consider a polygonal oriented link L in R3 (recall that any tame link is isotopic with a polygonal one) such that any edge is not perpendicular on the horizontal plane. An edge is named positive if its projection on the plane points counterclockwise viewed from the origin, and negative if not. By the chosen position of L, any edge is positive or negative. If all the edges are 14 positive, we take the projection on the plane (such a projection is named braided) and then cut the plane along a half-line from the origin; we obtain the desired braid. For example, in the picture below, we have two projection of the figure eight knot: the second is braided, while the first is not: Suppose we have a negative edge AB. If there is a point P on the z-axis such that the triangle P AB intersects L only along AB (such an edge is named accessible), then we can take a point C such that: -the triangle ABC cuts L only along AB -the triangle ABC contains P . By replacing AB with AC and CB as in the picture below, we arrive at a diagram for the same link but with fewer negative edges. If AB is not accessible, however any point in it is contained in an accessible sub-segment. By compactness of AB, it can be subdivided in a finite number of negative but accessible edges. These can be replaced by some positives with the previous method. We arrive again at a diagram with fewer negative edges. QED We turn now to the Vogel algorithm. Recall from the proof above the following: 15 Definition 2.5.2 A planar diagram D for an oriented link L is called braided if there exists a point O in the plane from which all edges of D are seen as counterclockwise oriented. The steps of the algorithm are intended, as in Alexander’s theorem, to transform the diagram into a braided one. There are two main operations to be considered. The first one is the smoothing, cf. the picture below: The second is the bending: Of course, a bending is in fact a Reidemeister II move, and so it does not change the link type. If we apply the smoothing operation at all crossings of D, we arrive at a disjoint union of what are called Seifert circles. Their number is denoted by n(D). Definition 2.5.3 Two Seifert circles are called incompatible if when considered as closed curves in S2 they are oriented as the boundary of the (conveniently oriented) annulus they bound. Let h(D) the number of pairs of incompatible Seifert circles. An important notion in the algorithm is the shadow | D | of D: Definition 2.5.4 The shadow of D is the same diagram with all crossings replaced by simple intersections. It is a 4-valent graph denoted by | D |. 16 Faces of | D | are the connected components in R2 \ | D |. Definition 2.5.5 A face of | D | is troubled if it has two opposite edges i.e. belonging to different and incompatible Seifert circles. The next two lemmas are the basis of the algorithm: Lemma 2.5.6 Let D0 obtained from D by a bending along two opposite edges. Then n(D0 ) = n(D) and h(D0 ) = h(D) − 1. Lemma 2.5.7 The shadow of a link diagram D has a troubled face iff h(D) > 0. The next lines are the beginning of the algorithm (cf. Prasolov and Sossinsky [23] pp. 58): DESTROY ALL CROSSINGS WHILE THERE IS A TROUBLED REGION DO A BENDING ALONG A PAIR OF OPPOSITE EDGES DESTROY ALL CROSSINGS END WHILE Applying the above ”program” we obtain a diagram without troubled regions. However it is not the end. The diagram is not necessarily braided. A last step is needed that can involve a so-called change of the infinity. (cf. figure below from [23] pp. 57) Lemma 2.5.8 An oriented link diagram D in R2 with h(D) = 0, can be transformed using the Reidemeister II and III moves into a braided one. 17 Remark 2.5.9 In fact, the hypothesis h(D) = 0 implies that D viewed in S2 is isotopic with a diagram which is braided in R2 (cf. lemma 2.6 in [14]). The fact that isotopic diagrams in S2 represent equivalent links is 2.1.2 from [14]. The only subtle point here is when the isotopy crosses the infinity, and it can be proved that from the planar point of view this can be obtained by Reidemeister II and III moves. The above lemma gives the second part of the algorithm: IF THE DIAGRAM IS BRAIDED STOP ELSE DO CHANGE THE INFINITY STOP. In fact by the invariance of n(D) under the bending operations and the fact that n(D) is also invariant by isotopy, we conclude that finally we obtain a braid representation with n(D) strands, the initial number of Seifert circles. 2.6 The fundamental group of the complement: braid group picture We have seen in Introduction, that the way to define a computable invariant passes through the fundamental group of the link/knot complement. The present and the next section are devoted to two methods for calculating this group. Here we shall use a braid presentation for the link L. If σ is an n-braid with closure L and σ̂ is the automorphism of the free group Fn in the letters x1 , ..., xn associated with σ, the following result due to Artin and Birman [22] gives a presentation of the fundamental group of the link complement: Theorem 2.6.1 π has a presentation of the form < x1 , ..., xn | x1 = σ̂(x1 ), ..., xn = σ̂(xn ) >. Remark 2.6.2 Any of the relations above is a consequence of the others and so we have a presentation with deficiency 1, a well known fact for link groups. 18 As an application of the above theorem let’s compute the Alexander polynomial of the Borromean link. The planar diagram is braided and a braid 3 for the Borromean link is (σ2 −1 σ1 ) . As consequence, the fundamental group π has the presentation: 3 3 π =< a, b, c | b = (σ2 −1 σ1 ) b, c = (σ2 −1 σ1 ) c >. The relations can be written as b = c−1 a−1 ca · b · a−1 c−1 ac and c = b−1 aba−1 · c · ab−1 a−1 b. With Fox calculus we obtain the following matrix M : −(xyz)−1 (y − 1)(z − 1) 0 −(xyz)−1 (y − 1)(x − 1) −1 −1 (xy) (y − 1)(z − 1) −(xy) (y − 1)(z − 1) 0 The elementary ideals are E1 = I(∆) where ∆ = (x−1)(y−1)(z−1) is the Alexander polynomial, and E2 = ((x−1)(y −1), (x−1)(z −1), (y −1)(z −1)). The characteristic varieties are V1 = V (∆) = {(x, y, 1), (1, y, z), (x, 1, z)} and V2 = {(x, 1, 1), (1, y, 1), (1, 1, z)}. 2.7 The fundamental group of the complement: Wirtinger picture The aim of this section is to describe Wirtinger presentation for a knot complement following mainly Rolfsen [25]. For a knot K, we begin with a plane presentation and a chosen orientation on it. We shall denote every connected component by αi and on it we shall chose an ortogonal vector xi such that: - it passes under αi - the ”frame” αi , xi agrees with a chosen orientation of the plane. The x0i s will be the generators of π in the sense that they represent based curves which go around the arcs αi0 s cf. figure below: 19 With respect to the types of crossings that can appear we can have the following situations: (figure from Rolfsen [25] pag. 57) The following four steps show us the imposed relation xk = xi+1 xk xi −1 for the left cross from above: 20 The algorithm above can be summarized in the following: Theorem 2.7.1 With the preceding notations π is presented with generators x0i s and relations ri0 s, where ri = xk −1 xi+1 xk xi −1 or ri = xk −1 xi −1 xk xi+1 . Moreover, any relation is a consequence of the others, i.e. the presentation of π is with deficiency 1. Exemple 2.7.2 For the left trefoil we have the following presentation π =< x, y | xyx = yxy >. Exemple 2.7.3 For the eight knot we have the following presentation π =< x1 , x2 , x3 , x4 | x2 x1 = x4 x2 , x2 x4 = x4 x3 , x1 x3 = x4 x1 >. 21 Remark 2.7.4 In fact the group of the figure eight knot can be presented by only two generators x, y with only one relation yx−1 yxy −1 = x−1 yxy −1 x. 22 Chapter 3 Alexander polynomial of knots and links The aim of this chapter, is to present several methods for the calculus of the Alexander polynomial for knots and links. As I mentioned in the Abstract, this invariant is the last, most tractable but coarser invariant on the road homeo type of the complement − π of the complement − Alexander invariant −Alexander polynomial. For a link L with q components in Σ an integral homology 3-sphere, we denote by X := Σ\L and G := π1 (X). We have the following: Proposition 3.0.1 G is a finitely presented group with def (G) > 0. We recall that ∆L (t1 , ..., tq ) ∈ Λ is the Alexander polynomial, AL the Alexander module, Vk (L) are the Alexander varieties and from 1.3.2, Corollary 3.0.2 We have: E1 (AL ) = (∆K ) if q = 1 (∆L ) · {I} if q > 1 Also, V1 (L) = V (∆L ) ⊆ T = (C? )q 3.1 Alexander polynomial via free Fox calculus From one point of view, Fox calculus is a highly efficient method for the calculus of the Alexander polynomial starting from a presentation of the 23 (fundamental) group. The main tool used is the notion of Fox derivation. Let G be a group, ZG its group ring and : ZG → Z the aditive agmentation morphism: (Σni gi ) = Σni . Definition 3.1.1 A (Fox) derivation is a map D : ZG → ZG such that: i) is aditive ii)D(w1 w2 ) = D(w1 )(w2 ) + w1 D(w2 ). A first interesting point is [3]: Theorem 3.1.2 For G = Fn , the free group on n letters x1 , ..., xn , for any 1 ≤ j ≤ n, there exist a unique derivation Dj such that Dj (xi ) = δi j . For a presentation of G < x1 , ..., xn | r1 , ..., rm >, we denote by ρ the composition ZFn → ZG → ZGab . Let’s consider the m × n matrix MG := [ρ(Dj (ri ))] with entries from ZGab . The central result for the Fox calculus is: Fox Theorem 3.1.3 MG is a presentation matrix for the Alexander module AG . Remark 3.1.4 We notice that the matrix above is a presentation matrix for the Alexander module AG and NOT for the Alexander invariant BG . Let’s compute the presentation matrix for AG for two traditional examples from [27]: Trefoil matrix 3.1.5 A presentation matrix for the trefoil is A = [t2 − t + 1; −t2 + t − 1], with t a generator of Gab = Z. Proof: For G we have the presentation < x, y | xyx = yxy >. Dx (xyx) = 1 + xy, Dx (yxy) = y so Dx (r) = 1 − y + xy. Also, by the same method Dy (r) = x − 1 − yx. After taking the image by ρ : ZF2 → ZG → ZZ, we obtain the mentioned result. QED Remark 3.1.6 In the calculus above we used the important fact that for a relation of the type a = b, for any Fox derivation, D(ab−1 ) = D(a) − D(b). 24 Eight knot matrix 3.1.7 A presentation matrix for the eight knot is A = [t − 3 + t−1 ; −t + 3 − t−1 ] Proof: As noted in 2.7.4 the fundamental group of the eight knot has a presentation of the form < x, y | yx−1 yxy −1 = x−1 yxy −1 x >. The images after abelianisation are: ρ(Dx (r)) = t − 3 + t−1 and ρ(Dy (r)) = −t + 3 − t−1 . QED From the two examples above, we obtain the Alexander polynomial for the trefoil and figure eight knots: ∆tref oil = t2 − t + 1 ∆f ig 3.2 eight = t2 − 3t + 1. Alexander polynomial via braid groups This section will provide a method to compute the Alexander polynomial of a link directly from an associated braid following Birman [1], Moran [22] and [14]. First of all we shall present what is named the Burau representation. Using it, there is a formula that produces the Alexander polynomial. For n ≥ 2 and i = 1...n − 1, we consider the n × n matrix Ui over the ring Λ = Z[t, t−1 ]: Ii−1 0 0 0 0 1−t t 0 Ui = 0 1 0 0 0 0 0 In−i−1 A simple calculus show that these matrices satisfy Ui Uj = Uj Ui for | i − j |≥ 2 Ui Ui+1 Ui = Ui+1 Ui Ui+1 for 1 ≤ i ≤ n − 2. and hence they produce a representation for the braid group Bn for n ≥ 2 in GLn (Λ). It is the Burau representation denoted by ψn . By convention, for n = 1 one consider the trivial representation B1 → GL1 (Λ). An important fact is that the Burau representations are compatible with the inclusions i : Bn ⊂ Bn+1 , which means that for any β ∈ Bn one has 25 ψn+1 (i(β)) = ψn (β) 0 0 1 Let’s consider for n ≥ 3 and 1 ≤ i ≤ n − 1 (n − 1) × (n − 1) matrices Vi defined by: −t 0 0 , V1 = 1 1 0 0 0 In−3 Vn−1 In−3 0 0 = 0 1 t 0 0 −t and for 1 < i < n − 1 Vi = In−2 0 0 0 0 0 0 0 0 1 t 0 0 0 −t 0 0 0 1 1 0 0 0 0 In−i−2 . Also, consider C the n × n matrix with 1 on and above the diagonal and 0, below. By ∗i we denote the row 1 × (n − 1) matrix with only 0’s if i < n − 1 or (0, ..., 0, 1) if i = n − 1. The Burau representations are in fact reducible ones and finally with the previous notations we can state the next theorem from [14]: Theorem 3.2.1 For 1 ≤ i ≤ n − 1 we have Vi 0 −1 C Ui C = ?i 1 As the Ui ’s verifies the braid relations, their conjugates by C verify the same relations and so we obtain what is called the reduced Burau representation ψn r : Bn → GLn−1 (Λ). For n = 2 it is defined by sending σ1 to the matrix −t. With Markov theorem in mind consider the following: Definition 3.2.2 A sequence of mappings fn : Bn → Z[s, s−1 ] is a Markov function if it is invariant under Markov moves. 26 Remark 3.2.3 In view of Markov theorem a Markov function produce a link invariant! Denote g : Λ = Z[t, t−1 ] → Z[s, s−1 ] the morphism which sends t → s2 and for β ∈ Bn , < β > its image under the morphism Bn → Z which sends all generators to 1. Also, consider the function fn : Bn → Z[s, s−1 ] for n ≥ 2 fn (β) = an (β) · g(det(ψn r (β) − In−1 )), where, the multiplication factor an is defined by an (β) = (−1)n+1 s −<β> (s−s−1 ) sn −s−n . For n = 1 we consider by definition f1 (B1 ) = 1. We arrived at the followings two fundamental results: Theorem 3.2.4 The above fn ’s defines a Markov function. Theorem 3.2.5 For a link L = β̂ for β ∈ Bn , fn (β) is the AlexanderConway polynomial. Exemple 3.2.6 For the right trefoil which is the closure of σ1 3 ∈ B2 , the above algorithm gives: f2 (σ1 3 ) = s2 +s−2 −1. For obtaining the normalized Alexander polynomial, we must consider the followings changes of variables: q √ s−1 − s → t − 1t . After this we arrive at the well known t + t−1 − 1. Exemple 3.2.7 For the figure eight knot depicted as below, 27 the corresponding braid is σ1 σ2 −1 σ1 σ2 −1 ∈ B3 . Using the two matrices V1 = V2 = −t 0 1 1 1 t 0 −t , and < σ1 σ2 −1 σ1 σ2 −1 >= 0, we arrive at the well known t + t−1 − 3. 3.3 Alexander-Conway polynomial via skein relations The present approach will produce an invariant for oriented links. At the end it will coincide with the previously defined versions of the Alexander invariant and in particular it is independent on the orientation in the knot case. We begin with the following: Definition 3.3.1 A Conway triple is composed by 3 oriented links which, outside a ball coincide, while in the ball they look like in the figure below 28 Definition 3.3.2 An Alexander Conway polynomial for links, is a function that assigns to every oriented link L, a polynomial ∇(L) ∈ Z[s, s−1 ] such that: 1. ∇(L) is invariant under isotopy, 2. it is 1 for the trivial knot, 3. for any Conway triple, we have ∇(L+ ) − ∇(L− ) = (s−1 − s)∇(L0 ). Remark 3.3.3 The example below is a Conway triple. It shows that if an Alexander Conway polynomial exists, it is 0 on all links which are union of a nonempty one with a trivial unlinked knot. In particular, on the trivial link with at last 2 components, it is 0. The main result of this section is: Theorem 3.3.4 An Alexander Conway polynomial exists, is unique and coincides with the fn defined using the reduced Burau representation. Proof: The proof, from [14], consists of two steps: we first prove uniqueness, and then, we show that the previously defined invariant satisfies the skein relation. Uniqueness: we need the following Definition 3.3.5 An oriented link diagram D is ascending, if it satisfies: 1. the link components can be indexed such that at every cross, the component with smaller index goes below that with a greater one, 2. any component has a base point (not a cross), such that if one move in the positive direction from that point, we meet every self-crossing first on the undergoing branch. 29 An ascending diagram represents always the trivial link. Suppose that we have two Alexander Conway functions and let ∇ their difference. We shall prove that ∇ = 0. ∇ is 0 on trivial knots and links and verify the skein relation. We proceed by induction on N -the number of crossings. For N = 0, we have a trivial knot/link an so the induction starts. Suppose it is true at level N . Let L a link presented with N + 1 crossings. At a cross, if we try to apply the skein relation, we obtain two other links: one , L0 with the same number of crossings but the other, L0 -the smoothed one, with only N crossings. By induction, ∇(L0 ) = 0 and so, ∇ is unchanged if we change any cross in L. But with these moves, we always reach the trivial link. So ∇(L) = 0 as we desired. Existence: we know that the fn defines a Markov function and hence a link invariant (under isotopy). It is easy that it is 1 on the trivial knot. We need to show the skein relation. Let n ≥ 2, 1 ≤ i ≤ n − 1 and α, β two braids in Bn . We have the following: FACT: ασi β, ασi −1 β and αβ are a Conway triple. The main point is that the proof of Alexander theorem (any link is the closure of a braid) show that any Conway triple is of this type. So we need to prove the skein relation only for triples as above: fn (ασi β) − fn (ασi −1 β) = (s−1 − s)fn (αβ). But fn is invariant under conjugation and σi is conjugated with σ1 , so we may assume i = 1. Also, by conjugation with α we can assume α = 1. So we need to prove: fn (σ1 β) − fn (σ1 −1 β) = (s−1 − s)fn (β). An intricate calculus, using the explicit form of fn (page. 117 from [14]) show the above relation. A last point to be verified is that the Alexander Conway invariant take values in Laurent polynomials, but it is an easy induction on the number of crossings that it is a polynomial in s−1 − s. QED Exemple 3.3.6 We apply this method to our favorite link: the left trefoil. 30 For the triple above, ∇(D− ) = ∇(D+ ) − z∇(D0 ), where by z we denote the s−1 − s from above. In the next step below, D0 became D− 0 and we have: ∇(D0 ) = ∇(D− 0 ) = ∇(D+ 0 ) − z∇(D0 0 ) = −z. So, the Alexanq q 2 √ √ der for the trefoil is: ∆(t) = ∇( t − 1t ) = 1 + ( t − 1t ) = t + t−1 − 1. 3.4 General properties of the Alexander polynomial This section collects the main properties of the Alexander polynomial for knots and links cf. [2] [18] [25]. KNOTS Theorem 3.4.1 For a knot K, ∆K (t) is symmetric (up to multiplication by units). 31 Theorem 3.4.2 For a knot, ∆K (1) = 1. If we denote by −K the same knot with orientation reversed and by K ∗ the mirror image (for a plane diagram this means changing all crossings) we have: Theorem 3.4.3 ∆−K = ∆K and ∆K ∗ = ∆K . In particular, the Alexander polynomial does not distinguish knots from their mirror images. Also, for factorizable knots we have: Theorem 3.4.4 ∆K1 ]K2 = ∆K1 · ∆K2 . LINKS We recall that a link is splittable if there is an embedded 2-sphere disjoint from L which separates some components of L from the others. A first property is: Theorem 3.4.5 For a splittable link L we have ∆L = 0. Theorem 3.4.6 For L a link with r components, the r-variables Alexander polynomial verifies the following Torres relations (up to multiplication by units): ∆L (t1 , ..., tr ) = ∆L (t1 −1 , ..., tr −1 ) and ∆L (t1 , ..., tr−1 , 1) = t1 l1 −1 ∆L0 (t1 ), t1 −1 t1 l1 ...tr−1 lr−1 ∆L0 (t1 , ..., tr−1 ), if r = 2 if r ≥ 3 where L = K1 ∪ ... ∪ Kr , L0 = K1 ∪ ... ∪ Kr−1 and li = lk(Ki , Kr ). 32 Chapter 4 Applications 4.1 Fox coloring The references for this section are [18], [24] and [26]. For an oriented link L ⊂ R3 , a planar diagram for it in R2 will be denoted by D. In general a Q-coloring of D is a map from the arcs of D (from the set of its connected components) to the set of ”colors” Q, such that certain conditions are satisfied at each cross. The crosses are defined to be positive/negative as in the following picture: Definition 4.1.1 A Fox or n-coloring is a Zn -coloring, with at least 2 colors such that at a colored cross we have: a + b = 2c (mod n). 33 Let Coln (D) = {n-colorings of D} and cn (D) =| Coln (D) |= number of ncolorings. With these notations we have the following results from [24] and [2]: Proposition 4.1.2 Coln (D) is an abelian group, cn (D) is a link invariant and for n = p, a prime number, cp (D) is a p-power, pν . Proposition 4.1.3 There is a bijection between Coln (D) ←→ Hom(G, D2n ) such that for c ∈ Coln (D) with c(si ) = ai the corresponding homomorphism is the one which sends si → ai b. Theorem 4.1.4 If L = K is a knot and n = p a prime number, then K has a p-coloring iff ∆K (−1) ≡ 0(mod p). Let n, d ≥ 2 integers, Q = Zn d and Φ the companion matrix of the degree d cyclotomic polynomial. Definition 4.1.5 A generalized Fox (or (n, d)) coloring is a Q-coloring such that, at every cross as below, the following relation holds: (Ci − Ck )Φ = Cj − Ck . In the above setting, Coln,d (D) is a Zn -module. More generally, for Q an abelian group and Φ ∈ Aut(Q) a fixed automorphism, we can define by the same method a generalized Fox Q-coloring by imposing the same relation at the crossings: (Ci − Ck )Φ = Cj − Ck . 4.2 2-Bridge knots and links In this section we will discuss about a particular class of knots/links named 2-bridge. 34 Definition 4.2.1 A 2-bridge link is one which has a planar diagram with exactly 2 minimums and 2 maximums. More precisely it is a particular type of closure of a 4-braid as in the picture bellow (cf. [2] pp. 25). These links are classified by 2 numbers (α, β) and denoted K(α, β) as in the following: Theorem 4.2.2 2-Bridge links are determined by a pair (α, β) such that: 1. 0 < β < α, β is odd and gcd(α, β) = 1 2. the 4-braid associated to the link is σ1 a1 σ2 −a2 ...σ1 am where m is odd and [a1 , ..., am ] are the quotients of the continued faction of β . α 3. K(α, β) is a knot for α odd and a link with 2 components for α even. Exemple 4.2.3 1. K(3.1) is the trefoil. 2. K(α, 1) is the (2, α) torus knot. 3. K(5, 3) is the figure eight knot. 4. K(8, 5) is the Whitehead link. The next step is to describe the fundamental group of K(α, β). For this we introduce: i·β i = (−1)[ α ] for i = 1, ..., α − 1. For a and b free variables we denote w = b1 a2 ...aα−1 if α is odd and w0 = b1 a2 ...bα−1 if α is even. We have the following theorem from [7]: 35 Theorem 4.2.4 If α is odd, the fundamental group of the complement of the knot K(α, β) has the following presentation: G(α, β) =< a, b | aw = wb >. If α is even the fundamental group of the 2-component link K(α, β) has the following presentation: G(α, β) =< a, b | aw0 = w0 a >. The following theorem from [13] concerns the Alexander Polynomial of 2-bridge knot: Theorem 4.2.5 The Alexander polynomial of the knot K(α, β) is: ∆K(α,β) (t) = 1 − t1 + t1 +2 − ... + t1 +...+α−1 Remark 4.2.6 For 2-bridge knots we have the following formula for the determinant: | ∆K(α,β) (−1) |= α. Proof: Using the formula for the Alexander polynomial in the previous theorem and using the fact that the mod 2 class of the power of t alternates we obtain +1 for each term in the sum. QED For two bridge links we have: Theorem 4.2.7 For α even, the linking number of the two components of K(α, β) is: α lk(K1 , K2 ) = 2 P 2j−1 . j=1 Also, using Torres relations we obtain: Theorem 4.2.8 Let l = lk(K1 , K2 ), the linking number of the two components in K(α, β). Then: ∆K(α,β) (−1, 1) = ∆K(α,β) (1, −1) = 36 1−(−1)l . 2 4.3 Metabelian quotients In this section we consider metabelian representation of knots groups. The main references are [18] and [11]. Definition 4.3.1 A group Γ is called metabelian if it is an extension of two abelian groups: 0→A→Γ→B→0 The last arrow is denoted by π : Γ → B. We shall assume that Γ is in fact a semi-direct product A o B. As a set Γ is A × B, but there is a morphism α : B → Aut(A), b → αb in terms of which, the multiplication in Γ is given by: (a, b)(a0 , b0 ) = (a + αb (a0 ), bb0 ). An equivalent formulation is that π has a section π 0 whose composition is the identity on B. Below are some examples: Exemple 4.3.2 1) Γ = D2n , the dihedral groups Zn o Z2 . 1 → Zn =< a >→ Γ → Z2 =< b >→ 1 where the the Z2 -action α on Zn is given by −1 → {a → −a}. In fact Γ has the following presentation: Γ =< a, b | an = b2 = 1, bab = a−1 >. 2) The metacyclic groups Γ = Zn o Zm , where m is a divisor of the order of Aut(Zn ). 3) Γ = Zn d o Zm with m a divisor of | Aut(Zn d ) |. For example the A4 group is Z2 2 o Z3 . For a group G we shall consider the set Rep(G, Γ). To obtain a description of it, we fix a morphism µ : G → B and we try to understand the set {λ : G → Γ | π ◦ λ = µ}, denoted by Repµ (G, Γ). Definition 4.3.3 A derivation φ : G → A is a map such that φ(gh) = φ(g) + g · φ(h), where the last multiplication, the G-module structure on A is given by ρ = α ◦ µ. Also, we denote by Zρ 1 (G, A) the set of derivations. 37 The description of Repµ (G, Γ), is given by the following: Proposition 4.3.4 There are bijections Repµ (G, Γ) ←→ Zρ 1 (G, A) and Repµ (G, Γ)/{inner auto0 s of Γ with el0 ts in A} ←→ Hρ 1 (G, A). The problem of the existence of morphisms from knot groups to metabelian groups is a difficult one. The following theorem, due to Fox, gives a criterion for the existence of such a morphism in the dihedral case, in terms of the Alexander polynomial of the knot: Theorem 4.3.5 For a knot K, there is a nontrivial morphism π1 (K) → D2p with p prime iff ∆K (−1) ≡ 0 mod p. An extension of the above theorem, was obtained by Matei and Suciu [19] in the following setting: for p, q prime numbers, denote by s the order of q mod p in Zp ? . The next lemma (pp. 485 in [19]) describes the metacyclic extensions Mp,qs : Lemma 4.3.6 1) There is an automorphism σ ∈ Aut(Zq s ) of order p. 2) All these automorphisms give isomorphic metacyclic extensions. 3) Aut(Zq s oσ Zp ) has sq s (q s − 1) elements. For G a finitely generated group, K a field and t : G → K? a character, the depth of t is: dK (t) = max{d | t ∈ Vd (G, K)}. Remark 4.3.7 We have 0 ≤ dK (t) ≤ l(G) (cf. [19] pp. 481), where l(G) is the minimal number of generators in a finite presentation of G. Let b the generator of Zp and Zq s viewed as the additive group of K = Fq (ξ), where ξ ∈ K? is a primitive p-th rooth. Then σ(b) can be identified with ξ ∈ Aut(K) and Zp with a sub-group in K? . So, Hom(G, Zp ) is a sub-set in the character torus Hom(G, K? ). The next theorem (pp. 487 in [19]) computes the number of (epi)morphisms from G to Mp,qs : 38 Theorem 4.3.8 The number of homomorphisms is: | Hom(G, Mp,qs ) |= Σ q sdK (ρ)+s , the sum being after all ρ ∈ Hom(G, Zp ). The number of epimorphisms is | Epi(G, Mp,qs ) |= Σ q s (q sdK (ρ) − 1), the sum being over all non-trivial ρ. Another interesting problem for a finitely generated group G and a finite one Γ is to compute the Hall invariants: δΓ (G) :=| Epi(G, Γ)/AutΓ |. In the above metacyclic setting we define Torsp,d (G, K) = {ρ ∈ Vd (G, K) of order exactly p}. The result below (cf. pp. 488 and 483 in [19]) compute the Hall invariants in terms of the number of torsion points in characteristic varieties of Mp,qs . Theorem 4.3.9 The Hall invariants are: δMp,qs (G) = p−1 s(q s −1) · Σ βp,d (q) (G) · (q sd − 1) the sum being over d ≥ 1 and βp,d (q) (G) = 1 · p−1 | Tors p,d (G, K)\ Tors p,d+1 (G, K) |. As an application we will compute the number of epimorphisms between the fundamental group of a 2-bridge knot and a dihedral group D2m . We know that G(α, β) =< x, y | xw = wy >, D2m =< a, b | am = 1; b2 = 1; bab = a−1 > and we have the following exact sequence: 1 → Zm → D2m → Z2 → 1. 39 A homomorphism f ∈ Hom(G, D2m ) is determined by (u, v, ) such that f (x) = au b , f (y) = av b . ∂ri Denote by Ψ : G → Zm , Ψ(x) = u and Ψ(y) = v, M = [( ∂x )] the Alexander j 0 matrix and ρ : G → Z2 the character such that the induced ρ : Gab = Z → Z2 is ρ0 (1) = (−1) . We have f ∈ Hom(G, D2m ) iff Ψ ∈ Derρ (G, Zm ). The last condition is equivalent with: M ((−1) )(u, v) ≡ 0 (mod m). Recall that the Alexander matrix for 2-bridge knot is M (t) = [−∆(t) ∆(t)] where ∆ is the Alexander polynomial. Denote by ΦH (G) =| Epi(G → H) | and σH (G) =| Hom(G → H) |. So the condition for f to be a homomorphism is equivalent to the following equation: ∆((−1) ) · (u − v) ≡ 0 (mod m). Case1 If = 0 ⇒ Imf ∼ Zl ⊆ Zm with l | m. But ΦZl (G) = ΦZl (Z) = φ(l) the last term being the Euler function. We know the formula: P φ(l) = m. l|m So, in this case we obtain m homomorphisms. Case2 If = 1, we have from the properties of Alexander polynomial for 2-bridge knots that ∆(−1) = α, so the equation: α · z ≡ 0 (mod m) where z := u − v. Denote by d = gcd(α, m), α = dα0 , m = dm0 with gcd(α0 , m0 ) = 1 ⇒ z ∈ {m0 , 2m0 , ..., dm0 } so gcd(α, m) solutions. For each solution z, we have u = z + v, v ∈ Zm so we obtain m pairs (u, v). It means that there are m · gcd(α, m) homomorphisms. From Case1 and Case2 we conclude that: σD2m = σD2m G(α, β) = m · gcd(α, m) + m. The dihedral group D2m has one subgroup which is isomorphic with Zl and m subgroups isomorphic with D2l , for l | m. l We will express the number of homomorphisms by the number of epimorphisms as follows: P m [ l ΦD2l + ΦZl ] . σD2m = l|m Applying Moebius inversion 40 ΦD2m = P l|m = P l|m m µ( ml )[σD2l l m µ( ml )[l l =m· P l|m − σZl ] = · gcd(α, l) + l − l] = µ( ml )gcd(α, l). After some computations it follows that: m · φ(m) if α ≡ 1(mod m) ΦD2m G(α, β) = 0 otherwise The Hall invariant for 2-bridge knot K(α, β) is: 1 if α ≡ 1(mod m) ΦD2m δD2m = |AutD2m | = 0 otherwise 41 Appendix A1. Manifolds and duality The main reference for this section is Hatcher [10]. Definition An n-manifold M is a Hausdorff topological space, where every point has a neighborhood homeomorphic with Rn . A compact manifold is named closed. As first examples of manifolds we have Rn , the spheres, the real or complex projective spaces, the open Moebius band, the genus g surfaces and the Klein bottle. For any commutative with unity ring R (usually it will be Z or Z2 ) and any point x ∈ M we have Hn (M, M \ x, R) ' R (by excision and homology of S n−1 ). An orientation of M is a function which for each x ∈ M assign a generator ex of Hn (M, M \ x, R), such that the following hold: Compatibility condition For any x there exist a chart neighborhood U and a generator eU of Hn (M, M \ U, R) ' R which for every y ∈ U goes over ey by the natural map Hn (M, M \ U, R) → Hn (M, M \ y, R). A first observation to made is that an orientation need not exists. An orientable manifold is one for which an orientation exists. For example on the Moebius band or the Klein bottle there is no orientation, but the spheres, any complex manifold and the real projective spaces of odd dimension are orientable. However any manifold admits a two-sheeted orientable covering. For closed orientable and connected n-manifolds we have the following: 42 Theorem The natural map Hn (M, R) → Hn (M, M \ x, R) is an isomorphism for all x ∈ M . Under the above conditions a generator of Hn (M, R) is named a fundamental or orientation class. For going further towards the Poincare duality theorem, we recall that for any space X there is a cap product defined at the chaincochain level, ∩ : Ck (X, R) × C l (X, R) → Ck−l (X, R), inducing a well defined, R-linear in each argument, map denoted also by ∩: Hk (X, R) × H l (X, R) → Hk−l (X, R). For σ : 4k → X and ϕ ∈ C l (X, R), σ ∩ ϕ is defined by: ϕ(σ(v0 , ..., vl ))σ(vl , ..., vk ), where v0 , ..., vk are the vertices of the standard k-simplex. With the above in mind, we arrive at the famous: Poincare duality theorem For M closed orientable with chosen fundamental class µ ∈ Hn (M, R), the map P D : H k (M, R) → Hn−k (M, R) defined by P D(x) = µ ∩ x is an isomorphism for every k. For orientable noncompact manifolds there is no fundamental class; however using cohomology with compact support there is a version of Poincare duality morphism H k c (M, R) → Hn−k (M, R) which is still an isomorphism. Another form of generalization of the Poincare duality is for manifolds with boundary. Definition An n-manifold M is a Hausdorff topological space, where every point has a neighborhood homeomorphic with Rn or with Rn + (the closed upper half space determined by the last coordinate). Points which by the chart homeomorphism go to a point with xn = 0 forms the boundary ∂M (an n − 1 manifold). A manifold with boundary is by definition orientable if M \ ∂M is. For a compact one, there is a (fundamental) class µ in Hn (M, ∂M, R) which restricts to the chosen orientation class at every point in M \ ∂M . Using product with µ we have: 43 Poincare-Lefschetz duality Let M compact orientable n-manifold with his boundary decomposed as A ∪ B, where A, B are n − 1-manifolds with common boundary A ∩ B. Then, P D : H k (M, A, R) → Hn−k (M, B, R) is an isomorphism. In particular: for A = Φ and B = ∂M , H k (M, R) ' Hn−k (M, ∂M, R); for B = Φ and A = ∂M , H k (M, ∂M, R) ' Hn−k (M, R). A2. Immersions, embeddings, isotopy In the preceding section manifolds were only topological. In this one, sometimes a manifold will be differentiable which means that it has an open covering with chart domains, such that the transition functions are differentiable of class C ∞ . Definition A differentiable map f : M → N between smooth manifolds is an immersion if at every point x ∈ M the differential is injective. Definition A continous map f : M → N between topological spaces is a topological embedding if it is injective and homeo on its image.(the image having the subspace topology) Note that an injective map needs not always be a topological embedding; for example the figure eight in the plane is the injective image of any open interval of R, without being homeo with it. However we have the following easy result: Theorem An injective map from a compact space into a Hausdorff one is a topological embedding. In the differentiable case we have: 44 Definition A differentiable map f : M → N between smooth manifolds is a differentiable embedding if it is an immersion and a topological embedding. We remark that there are differentiable topological embeddings which are not differentiable embeddings:f (x) = x3 from R → R is an obvious example. The third subject of this section is the notion of isotopy. Definition Two topological embeddings f, g : X → Y are isotopic if they are homotopic through embeddings (i.e. at each floor the corresponding map is an embedding). A related notion is the ambient isotopy: Definition Two topological embeddings f, g : X → Y are ambient isotopic if there exists F : Y × [0, 1] → Y such that F0 = id, Ft is a homeo and F1 ◦ f = g. It is clear that ambient isotopy is a stronger relation than isotopy and in fact it is the equivalence we are working with in knot/link theory. However in the differentiable setting with compact source these are in fact equivalent cf. Hirsch [12]. If the source is not compact one can consider a line in R3 and a line modified somewhere by a trefoil knot. These are isotopic but not ambient isotopic because their complements have not the same π1 . For knots/links, there is also the following third notion introduced by Kauffman [15], but we will not be concerned with it in this thesis. Definition Two link diagrams are regular isotopic if they are connected through type II and III Reidemeister moves. A3. Rings and orders The aim of this section is to describe the theory of orders and Fitting ideals of modules, using as main references [18], [4], [6], [22], [25] and [27]. R will always denote a commutative, integer, unitary, unique factorization ring (UFD for short). Sometimes it will be even a principal ideal domain (PID for short). M is a finitely presented R-module. A presentation of M is an exact sequence Rm → Rn → M → 0 45 where m, n positive integers. Using standard basis in the free modules above, the presentation is encoded in an m × n matrix A with elements from R. We have the following fundamental definition: Definition For natural k ≥ 0 the k-elementary ideal (or Fitting or determinantal) is the ideal in R denoted by Ek (M )generated by all (n − k) × (n − k) minors from A.(by convention they are 0 if n − k > m and all R if n − k ≤ 0) As far as every minor is a linear combination of sub-minors, the elementary ideals forms an ascending sequence. We have the following theorem: Theorem The elementary ideals are invariants of the module M and do not depend on the chosen presentation. Now, R being an UFD and M finitely presented, if ∆k (M ) is the greatest common divisor of elements in Ek (M ), then ∆0 (M ) is called the order of M and we have: Lemma ∆k+1 (M ) | ∆k (M ) for k ≥ 0. Exemple Suppose M = Rr ⊕ (pR1 ) ⊕ ... ⊕ (pRs ) is a finitely generated module over a PID (eg. R = K[t±1 ], where K is a field ). Then: if i < r 0 1 if i ≥ r + s ∆i (M ) = pi−r+1 ...ps if r ≤ i < r + s Remark Through this thesis, R will be K[t1 ±1 , ..., tq ±1 ], for K = Z or C, the Laurent polynomials ring; it will be denoted by Λ ⊗ K. For E an ideal in R as above, V (E) is the reduced variety defined by E in S = Spec(R). For M a finitely generated R-module supp(M ) := V (ord M ). Moreover, we have the following: Definition The k th -support variety of M is Vk (M ) := V (Ek−1 (M )) ⊆ S. In particular V1 (M ) = V (ord M ). 46 Exemple For R = Λ ⊗ C = C[t1 ±1 , ..., tq ±1 ], S = Spec(R) = (C? )q =: T. For an ideal E 6= 0 in Λ ⊗ C and ∆ = gcd(E) the generator of the smallest principal ideal containing E, T ⊇ V (E) ⊇ V (∆) and we have: Lemma V (∆) = W1 (E) := the union of all codimension 1 irreducible components of V (E). 47 Directions for further study As we had seen along the above pages, the Alexander modules/polynomials are strong tools for the study of topological and algebraic properties of 3dimensional complements. For the future, I think that many directions are possible. First of all, a further study concerning the relation between Alexander varieties and metabelian representations should be very interesting. Secondly, the modern direction toward the ”twisted” world would be a natural next step. Last but not least, an interesting route is the study of the Alexander ideas in other contexts like complements of projective hyper-surfaces. All these directions are under current development across the world. 48 Bibliography [1] J. S. Birman, Braids, Links, and Mapping Class Groups, Princeton University Press, 1975. [2] G. Burde, H. Zieschang, Knots, De Gruyter Studies in Mathematics, no. 5, 2002. [3] R. H. Crowell, R. H. Fox, Introduction to Knot Theory , Dover Books on Mathematics, 2008. [4] A. Dimca, S. Papadima, A. Suciu, Alexander polynomials: essential variables and multiplicities, Int. Math. Research Notices vol.(2008), no. 3, Art. ID rnm119, 36 pp. [5] D. Eisenbud, W. D. Neumann, Three-Dimensional Link Theory and Invariants of Plane Curve Singularities, Princeton University Press (1986). [6] R. A. Fenn, Techniques of Geometric Topology, London Mathematical Society Lecture Note Series, Cambridge University Press, (1983). [7] S. Fukuhara, Explicit formulae for two bridge knot polynomials, J.Aust. Math. Soc. 78 (2005), pp. 149-166. [8] C. Gordon, J. Luecke, Knots are determined by their complements, Bull. AMS.(1989), vol. 20, no. 1, pp. 83-87. [9] C. Gordon, J. Luecke, Knots are determined by their complements, JAMS.(1989), vol. 2, no. 2, pp. 371-415. [10] A. Hatcher, Algebraic topology, Cambridge University Press, (2001). [11] P. J. Hilton, U. Stammbach, A course in homological algebra, GTM Springer, Second Edition, (1997). 49 [12] M. W. Hirsch, Differential topology, GTM, vol. 33, Springer, (1976). [13] J. Hoste, P. D. Shanahan, Twisted Alexander polynomials of 2-bridge knots, Journal of Knot Theory and Its Ramifications Vol. 22, No. 1 (2013) DOI: 10.1142/S0218216512501386 [14] C. Kassel, V. Turaev, Braid groups, Springer, (2008). [15] L. H. Kauffman, An invariant of regular isotopy, Trans. of the AMS. (1990), vol. 13, no. 2, pp. 417-471. [16] A. Kawauchi, A survey of knot theory, Birkhauser, (1996). [17] V. Manturov, Knot theory, CRC Press, 2004. [18] D. Matei, Handwritten notes. [19] D. Matei, A. Suciu, Hall Invariants, homology of subgroups and characteristic varieties, International Math. Research Notices 9 (2002),465-503 [20] D. Matei, A. Suciu, Counting homomorphisms onto finite solvable groups, J. of Algebra 286 (2005), 161-186. [21] J. W. Milnor, Infinite cyclic coverings, Conference on the Topology of Manifolds (Michigan State Univ., E. Lansing, Mich., 1967), Prindle, Weber & Schmidt, Boston, Mass., 1968, pp. 115-133. [22] S. Moran, The Mathematical Theory of Knots and Braids: An Introduction , North Holland, 2000. [23] V. V. Prasolov, A. B. Sossinsky, Knots, Links, Braids and 3-Manifolds: An Introduction to the New Invariants in Low-Dimensional Topology , Translations of Mathematical Monographs AMS, 1996. [24] J. H. Przytycki, The Trieste look at knot theory, arxiv:1105.2238v1. [25] D. Rolfsen, Knots and Links, University of British Columbia - AMS, 1976. [26] D. S. Silver, S. G. Williams Generalized n-coloring of links, Knot Theory, Banach Center Publ., vol. 42, pp.381-394. [27] V. Turaev, Introduction to combinatorial torsion, Birkhauser, 2001. 50 [28] http://www.tlhiv.org/rast2vec/. 51